SUMMARY: strange problem with pnmd from suncluster

Hi @ll,

thanx to our contact at SUN for the right solution:

add the following lines to /etc/system:

     set ce:ce_reclaim_pending=1
     set ce:ce_taskq_disable=1

and apend this to /platform/sun4u/kernel/drv/ce.config:

     interrupts=1;

Bye,
Marcel


Original message:

> Hi managers,
>
> I've got a problem with our suncluster 3.0 ...
>
> The cluster's running on two Fire-480-nodes. On both nodes I defined a
> nafo group "nafo0" with only one adaptor "ce0" (due to time and gardware
> lack).
>
> Now I get the following in /var/adm/messages on _both_ nodes:
>
> ++++++++++++++++++++++
>
> Aug  5 08:28:53 fix Cluster.PNM: [ID 164168 daemon.error] PNM: nafo0:
> state transition from DOWN to OK on ce0
> Aug  5 08:33:27 fix Cluster.PNM: [ID 143622 daemon.notice] PNM: adapter
> ce0 is faulty
> Aug  5 08:33:27 fix Cluster.PNM: [ID 164168 daemon.error] PNM: nafo0:
> state transition from OK to DOUBT on ce0
> Aug  5 08:33:27 fix Cluster.PNM: [ID 164168 daemon.error] PNM: nafo0:
> state transition from DOUBT to DOWN on ce0
> Aug  5 08:35:48 fix Cluster.PNM: [ID 143622 daemon.notice] PNM: adapter
> ce0 is ok
> Aug  5 08:35:48 fix Cluster.PNM: [ID 164168 daemon.error] PNM: nafo0:
> state transition from DOWN to OK on ce0
> Aug  5 08:37:10 fix Cluster.PNM: [ID 143622 daemon.notice] PNM: adapter
> ce0 is faulty
> Aug  5 08:37:10 fix Cluster.PNM: [ID 164168 daemon.error] PNM: nafo0:
> state transition from OK to DOUBT on ce0
> Aug  5 08:37:10 fix Cluster.PNM: [ID 164168 daemon.error] PNM: nafo0:
> state transition from DOUBT to DOWN on ce0
> Aug  5 08:39:19 fix Cluster.PNM: [ID 143622 daemon.notice] PNM: adapter
> ce0 is ok
> Aug  5 08:39:19 fix Cluster.PNM: [ID 164168 daemon.error] PNM: nafo0:
> state transition from DOWN to OK on ce0
>
> +++++++++++++++++++++++++++++++
>
> This 'nafo-pingpong' does not occur on our E420-cluster (two nodes, too)
> with hme-devices. Due to the fact that I saw a Fire-480 first time last
> week I'm not sure what causes this. (The node itself is reachable over
> that interface permanently, only the clustersoftware does not think so
> :-( and because of that several resource groups often switch primaries.)
>
> Should anybody have an idea what _should be_ wrong, I could send a
> tracefile of pnmd...
>
> Of course I'll summarize.
>
>
> Thanx a lot for your help,
>
> Marcel
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers